LETTER Doi:10.1038/Nature14452

LETTER doi:10.1038/nature14452

TH17 cells transdifferentiate into regulatory T cells during resolution of inflammation

Nicola Gagliani1, Maria Carolina Amezcua Vesely1*, Andrea Iseppon1*, Leonie Brockmann2,HaoXu1, Noah W. Palm1, Marcel R. de Zoete1,3, Paula Licona-Limo´n1{, Ricardo S. Paiva1, Travers Ching4, Casey Weaver5, Xiaoyuan Zi6{, Xinghua Pan7, Rong Fan6, Lana X. Garmire4, Matthew J. Cotton8, Yotam Drier8, Bradley Bernstein8, Jens Geginat9, Brigitta Stockinger10, Enric Esplugues11, Samuel Huber21 & Richard A. Flavell1,31

Inflammation is a beneficial host response to infection but can con- Despite these similarities, it is unclear if TH17 cells transiently co- tribute to inflammatory disease if unregulated. The TH17 lineage of express a limited number of genes that are typically associated with Thelper(TH) cells can cause severe human inflammatory diseases. regulatory CD4 T cells, or if they can undergo genetic and functional These cells exhibit both instability (they can cease to express their reprogramming resulting in transdifferentiation from one TH type to signature cytokine, IL-17A)1 and plasticity (they can start expressing another. cytokines typical of other lineages)1,2 upon in vitro re-stimulation. To track TH17 cell fate towards regulatory states in vivo, we crossed However, technical limitations have prevented the transcriptional IL-17A fate reporter mouse (IL-17ACRE 3 Rosa26 STOPfl/fl YFP YFP Katushka eGFP RFP profiling of pre- and post-conversion TH17 cells ex vivo during (R26 ))1 with IL-17A IL-10 Foxp3 triple reporter 1 immune responses. Thus, it is unknown whether TH17 cell plasticity mouse model9,19. We call the resulting mouse model Fate merely reflects change in expression of a few cytokines, or if TH17 (Methods, Extended Data Fig. 1a, b) in which, cells that have prev- cells physiologically undergo global genetic reprogramming driving iously expressed high level of Il17a, delete the stop cassette preceding their conversion from one T helper cell type to another, a process R26YFP and are permanently marked by YFP expression. This enabled 1 known as transdifferentiation3,4. Furthermore, although TH17 cell us to test if YFP cells express IL-17A, IL-10 and Foxp3 ex vivo without instability/plasticity has been associated with pathogenicity1,2,5,itis in vitro restimulation. unknown whether this could present a therapeutic opportunity, In steady state TH17 cells are mainly in the small intestine due to the whereby formerly pathogenic TH17 cells could adopt an anti-inflam- presence of segmented filamentous bacteria (SFB)12. Among intestinal matory fate. Here we used two new fate-mapping mouse models CD4 T cells approximately half (48% 62.7, n 5 18) of the cells that had 1 to track TH17 cells during immune responses to show that CD4 expressed IL-17A no longer expressed this cytokine. We call these cells Katushka– 1 T cells that formerly expressed IL-17A go on to acquire an anti- exTH17 cells (IL-17A YFP ). Some (4.3% 6 0.3, n 5 18) eGFP inflammatory phenotype. The transdifferentiation of TH17 into reg- intestinal exTH17 cells expressed IL-10 , and some (1% 60.2, RFP ulatory T cells was illustrated by a change in their signature tran- n 5 18) of them were Foxp3 positive (Fig. 1a, b). ExTH17 IL- eGFP1 scriptional profile and the acquisition of potent regulatory capacity. 10 cells were distinct from TH1, TH2andTH17 cells since they Comparisons of the transcriptional profiles of pre- and post- expressed trace amounts of IFN-c, were negative for IL-4, and conversion TH17 cells also revealed a role for canonical TGF-b expressed low levels of RORct and CCR6 respectively (Extended signalling and consequently for the aryl hydrocarbon receptor Data Fig. 1c–e). Finally, to test if the presence of TH17 and conse- (AhR)inconversion.Thus,TH17 cells transdifferentiate into quently exTH17 was due to SFB, we treated the mice with vancomycin; regulatory cells, and contribute to the resolution of inflam- both populations were reduced (Fig. 1a, b). Thus under homeostatic mation. Our data suggest that TH17 cell instability and plasticity conditions, intestinal TH17 cells lose IL-17A expression and a fraction is a therapeutic opportunity for inflammatory diseases. of these exTH17 cells express regulatory features but not characteristic TH17 cells are characterized by secretion of IL-17A, expression of signatures of TH1, TH2 and TH17 cells. chemokine receptor CCR6 and transcriptional factor RORct6. Their We next analysed TH17 cell plasticity during a self-limiting inflam- 1 pathogenicity is limited by Foxp3 TReg and T regulatory type 1 (TR1) matory response induced by the injection of anti-CD3 monoclonal 7,8 1 8 cells . Foxp3 TReg cells are characterized by the transcription factor antibody . Intestinal TH17 cell expansion was followed by increased eGFP Foxp3, whereas TR1 cells secrete high levels of the anti-inflammatory exTH17 cells expressing high IL-10 (Fig. 1a, b), although few IL-10 and express cell-surface markers CD49b and LAG-3 (refs 7, 9–11). (2% 60.2, n 5 8) of these cells co-expressed IL-10eGFP and Foxp3RFP. 1 1 Although TH17, Foxp3 TReg and TR1 cells are functionally distinct The low number of Foxp3 exTH17 cells prevented, at the time, further subsets, they share some features. They are abundant in the intestine, studies on these cells. eGFP1 their differentiation is promoted by transforming growth factor As exTH17 IL-10 cells resembled TR1 rather than TH17 cells, b (TGF-b)12, and both TH17 and TR1 cells express CD49b and high we examined them for cell-surface markers that identify TR1 and levels of the transcription factor AhR9,13. Moreover TH17 cells can TH17 cells – LAG-3 (ref. 9), and CCR6 (ref. 12). A high percentage eGFP1 transiently co-express RORct with Foxp3 (refs 14, 15), and IL-17A of exTH17 IL-10 cells were LAG-3 positive but CCR6 negative. with IL-10 (refs 10, 16–18). Interestingly, in contrast to chronically activated and colitogenic TH17

1Department of Immunobiology, School of Medicine, Yale University, New Haven, 06520, USA. 2Medizinische Klinik und Poliklinik, Universita¨tsklinikum Hamburg-Eppendorf, Hamburg 20246, Germany. 3Howard Hughes Medical Institute, Yale University School of Medicine, New Haven, Connecticut 06520, USA. 4University of Hawaii Cancer Center, Manoa 96813, USA. 5Department of Pathology, University of Alabama at Birmingham, Birmingham, Alabama 35294, USA. 6Department of Biomedical Engineering, Yale University, New Haven, 06520, USA. 7Department of Genetics, Yale University School of Medicine, New Haven, Connecticut 06520, USA. 8Howard Hughes Medical Institute and Department of Pathology, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114, USA. 9Istituto Nazionale di Genetica Molecolare ‘‘Romeo ed Enrica Invernizzi’’, Milan 20122, Italy. 10Division of Molecular Immunology, MRC National Institute for Medical Research, Mill Hill, London NW7 1AA, UK. 11Immunology Institute, Mount Sinai School of Medicine, Icahn Medical Institute, New York, New York, USA. {Present addresses: Departamento de Biologıá Celular y del Desarrollo, Instituto de Fisiologıá Celular, Universidad Nacional Autońoma de Me´xico, D.F. Me´xico 04510, Me´xico (P.L.-L.); Department of Cell Biology, Second Military Medical University, Shanghai 200433, China (X.Z.). *These authors contributed equally to this work. 1These authors jointly supervised this work.

cells, which are LAG-3 negative9,TH17 cells expressed low levels of Methods, Extended Data Fig. 4a–c) in which TH17 cells become LAG-3 cells during this self-limiting immune response, supporting the YFP1 only after tamoxifen treatment (Fig. 1d and Extended Data idea of an ongoing maturation towards a TR1 cell phenotype. As Fig. 4d). Through the use of the iFate model we observed that TH17 expected9 CD49b was equally expressed among the three populations still convert into TR1 cells, specifically during the immune response eGFP1 (Extended Data Fig. 2a–c). Like TR1, exTH17 IL-10 cells (anti-CD3 monoclonal antibody + tamoxifen; Fig. 1d). exT 17 expressed low levels of RORct (Extended Data Fig. 2d) and only trace We next examined whether TR1 H cells undergo transcriptional levels of characteristic TH1andTH2 genes (Extended Data Fig. 2e). In reprogramming during their conversion from TH17 into TR1 cells. We exT 17 conclusion, during a self-limiting response, some exTH17 cells sequenced the transcriptome of intestinal TR1 H cells and com- exT 17 resemble TR1 (hereafter named TR1 H ) rather than TH17 cells. pared it to the transcriptomes of bona fide TR1andTH17 cells isolated We next determined TH17 fate during a non-resolving immune from the same mice (Extended Data Fig. 5a, b). As controls, we used 2 1 1 1 response. DNIL-10R transgenic mice have an impairment in IL-10R exTH17 IL-10 , Foxp3 TReg and Foxp3 Treg IL-10 cells, again signalling in CD4 T cells and when treated with anti-CD3 the inflam- H 8 Colour key mation does not resolve, but leads to T 17-associated mortality .We abTH17 relevant genes Cytokines found that in Fate DNIL-10R mice, in which immune response cannot 0.5 0.8 be terminated, exTH17 cells tend to acquire a TH1-like phenotype Value exTH17 H TR1 rather than a TR1-like phenotype (Extended Data Fig. 3a–c). TR1exT 17 exTH17 The presence of TR1 cells under steady state conditions sug- TR1 TR1 exTH17 gests two models to explain TR1 cell formation during an Foxp3IL-10+ Foxp3IL-10+ immune response. First, TH17 might convert to TR1 cells in steady Foxp3 ExTH17 state and during an immune response, such cells expand (Fig. 1c). ExTH17 TH17 Alternatively, TH17 cells might convert to TR1 cells over the course Foxp3 of the response (Fig. 1c). To distinguish these possibilities, we gener- TH17 eGFP 1 17 1 17 17 17 R H 17 17 R H H H T H H T IL-10+ ated a tamoxifen inducible IL-17A fate mouse model ((iFate) T IL-10+ T exT Foxp3 exT Foxp3 1 ExT 1 ExT R R T T Foxp3 Foxp3 c in vivoo a b in vitro TH17 ** generated ExTH177 generatedted 22.2%.2% 5% 0.9% *** eGFP 6 ** 100,000 ***

) eGFP Katushka IL-17A TR1exTH17 Fate+ 10 –/– IL-10 Foxp3RFP Rag1 4 10,000 Pathogenic TH17 TR1 IL-17A Steady state

11.6%.6% 17 cells H

ExTH17 cells 17 cells (%) 2 1,000

H Colitis? ExT 00.1%.1% (numbers, Log pTH17 +++ + ExT d 0 100 exTH17 TH17 ExTH17 TR1 TR1 ***1,500 * vancomycin

Steady state 00.07%.07% 40 1,000 cells s exTH17 exTH17 g

H TR1 Foxp3 17 EExTxT 1177 ccellsells cells (%) g H t 17 g H 224%4% 35% 2% 20 exT s 1 500 t exT R t s 1 T R IL-10 eGFP (MFI) IL-10 eGFP T 0 0

44%% Steady state * Inflammation ** ** Vancomycin EExTxTH1177 ccellsells C * YFP Foxp3RFP Inflammation C C * c d 0% (YFP+) First model Second model ef*** * ** * Steady State Imm.response Imm. response ***** * ** *** Katushka 12 120 TH17 TH17

IL-17A 110 TR1exTH17 TR1exTH17 9 Steady state

20% (YFP+) 100 46.3% 6 666%6% CD90.1

Colits score 90

IL-10 3 Proliferation 80 + tamoxifen Inflammation 330%0%

0 Percentage of initial body weight 70 Conversion No conver. YFP CD4 pTH17 + ++ + pTH17 + ++ +

17 1 17 1 17 17 H R 17 17 H R H H T H H T T exT T exT Figure 1 | TH17 cells lose IL-17A and acquire IL-10 in vivo.a,Flow 1 1 ExT R ExT R cytometric analysis of small intestinal CD41 T cells. Steady state, vancomycin- T T exT 17 treated, or treated with anti-CD3 monoclonal antibody (inflammation) Figure 2 | TR1 H cells have a similar gene expression profile and function 1 depicted. b, Number and frequencies of exTH17 cells (gated on CD4 Tcells) compare with TR1cells.a, b, Correlative heatmaps based on the expression of exT 17 and TR1 H cells (gated on exTH17 cells) are cumulative of two and three TH17 related genes (n 5 97) (a) and cytokine genes (n 5 191) (b). The indicated independent experiments respectively. IL-10 mean fluorescence intensity (MFI) cell populations were isolated from the small intestine of 10 anti-CD3 treated 1 data (n 5 3 biological replicates) of one representative experiment out of three Fate mice from two independent experiments. c, Pathogenic (p)TH17 were are shown. Mean 6 s.e.m.; *P # 0.05, **P # 0.005, ***P # 0.0005 by ANOVA differentiated in vitro and then injected alone or in combination with the 2/2 exT 17 (Bonferroni’s multiple comparison test) or by t-test for percentage (Mann- depicted populations into Rag1 mice. TH17, exTH17, TR1 H and TR1 exT 17 2 1 Whitney U-test, two tailed) and MFI (paired t-test, two-tailed) of TR1 H cells. cells (YFP ) were isolated from the small intestine of Fate mice treated with exT 17 c, Hypotheses: first model, expansion of pre-existing TR1 H ; second model, anti-CD3 mAb. d, Endoscopic and histological pictures. Scale bars: 200 mm. conversion of TH17 cells expanded/induced over the course of immune Endoscopy pictures show stool inconsistency (s), increased mucosal granularity response. d, Flow cytometric analysis of small intestinal CD41 T cells isolated (g), lack of translucency (t) and bleeding (arrow). The histological pictures from iFate mice in steady state, and upon anti-CD3 monoclonal antibody and show oedema (**), inflammation (*) and crypt loss (C). e, f, Endoscopic colitis tamoxifen (Inflammation 1 tamoxifen) treatment. Frequencies of YFP1 cells, score (e) and percentage of initial body weight (f). Each dot represents one 1 1 TH17 and exTH17 (gated on YFP cells), and of IL-10 cells (gated on exTH17) mouse. Mean and s.e.m. are indicated. *P # 0.05, **P # 0.005, ***P # 0.0005 are representative of two experiments (n 5 6biologicalreplicates). by ANOVA (Tukey’s multiple comparison test).

exT 17 isolated from the same mice. To determine the relatedness of these (MOG)) were TR1 H cells. We also observed that MOG-specific subsets, we performed hierarchical clustering based on the expression exTH17 cells express IL-10 to a greater extent than non-MOG-specific of literature-search curated TH17-relevant genes20,21 (Supplementary exTH17 cells (Fig. 3f–i). Likewise, in iFate mice TH17 cells labelled Table 1). Clustering analyses revealed first that TR1 are a distinct T-cell during EAE onset converted to TR1 cells (Extended Data Fig. 6a, b). 1 subset, as different to TH17 or Foxp3 TReg cells as the latter two cell Thus pathogenic autoantigen-specific TH17 cells can convert to exT 17 exT 17 types differ from each other and second that TR1 H cells cluster TR1 H cells. together with TR1 rather than with TH17 cells (Fig. 2a). We also per- We next addressed if TH17 cells convert to TR1 cells during an formed cytokine-restricted cluster analysis (Supplementary Table 2) immune response, which physiologically promotes host tolerance exT 17 showing that TR1andTR1 H cluster together (Fig. 2b). Thus, con- to infection23. Nippostrongylus brasiliensis infection elicits a type-2 version of TH17 into TR1 cells is determined and/or followed by a immune response that drives worm expulsion24.InadditiontoTH2 reprogramming of the TH17-relevant transcriptional profile, a process cells, TH17 and TR1 also expand in response to N. brasiliensis9 and previously described as transdifferentiation3,4. while IL-17 contributes to tissue damage, IL-10 prevents tissue exT 17 As a final test of whether TR1 H cells had completed their func- damage24.WethereforeaskedifTH17 cells convert to TH2cells25 tional trans-differentiation from TH17 into regulatory TR1 cells, we or to TR1 cells during N. brasiliensis infection. During the primary exT 17 used a TH17 cell-mediated colitis model8. We found that TR1 H immune response to N. brasiliensis, TH17 cells lost IL-17A express- cells had completed their functional reprogramming since they pre- ion and some showed a TH2phenotype25. However, when we vented TH17 cell-mediated colitis (Fig. 2c–f). re-infected the mice with N. brasiliensis we observed TH17 conver- TH17 cells can also differentiate into TH1-like cells in a TH17 sion into a TR1 cell-phenotype (Fig. 3j–l). We confirmed these find- mediated-mouse model for multiple sclerosis (EAE) (Fig. 3a–e)1,22. ings in iFate mice (Extended Data Fig. 6c–e). Thus, TH17 cells can We next wondered whether autoimmune-derived TH17 cells would become TR1 cells during the secondary response to N. brasiliensis be able to acquire a regulatory fate. Of note, after anti-CD3 monoclonal infection and this may limit potentially destructive type 1 immune antibody treatment, which can block EAE development17, a fraction of responses. exTH17 cells acquired IL-10, not IFN-c (Fig. 3a–e). Thus, some Finally we asked if TH17 conversion to TR1 cells occurs in exTH17 cells, which developed during an autoimmune response, can response to acute bacterial infection. S. aureus causes sepsis in human still convert into TR1 cells. and patients with TH17 associated gene deficiency suffer recurrent 1 Furthermore, encephalitogenic TH17 cells can be recruited to the S. aureus infections26. Intravenously S. aureus infected Fate mice small intestine after anti-CD3 monoclonal antibody treatment17.In (Extended Data Fig. 7a–c) and iFate mice (Extended Data Fig. 7d, e) our current studies some intestinal T cells specific for the disease- accumulated TH17 cells in the small intestine, as we previously showed17, driving antigen of EAE (myelin oligodendrocyte glycoprotein and in both models TH17 cells acquired a TR1 cell phenotype.

abcd 4 Day 10 Day 35 Control Anti-CD3 treated 15 **

3 62% 10% 0.7% 7.5%

Anti-CD3 mAb 17 (%) 10 H Katushka eGFP 2 Katushka 5

38% 90% IL-10 1 1 exT R IL-17A IL-17A T EAE clinical score 0 0 0 10203040 YFP CD4 Days after immunization YFP Control Anti-CD3

Control Anti-CD3 treated e f gh100 86% 64% Katushka eGFP 1,975 **

CD4 ±111 MFI 80 14% IL-10

0.34% 50.9% 8.12% IL-17A

17 (%) 60 MOG-positive H 40 1 exT

MOG-tetramer R T 80.5% 42% MOG-negative 700 20 ±91 MFI l ** 19.5% 0 k 10 days 7 days 6 ** after first inf. after second inf.

4 YFP CD4 MOG positive IL-17A MOG negative i Control Anti-CD3 treated 17 (%)

H 2 0.05%05% 0.24% 2% 6% 0.3% 26% 5% ExT 0 eGFP rl

Ct IL-10 YFP First inf. 40 Second* inf. 16% 10% 0.0% 0.0% 4.96% 0.83% 20.32% 2.27% IFN-γ

eGFP 30 10 - j L 20 First N. Brasiliensis IL-10 I Second N. Brasiliensis 17 (%)

H infection infection 10

1 exT Day 0 Day 10 Day 25 Day 32

R 0 T IL-4IL-4 . Analysis Analysis Ctrl First inf Second inf.

exT 17 Figure 3 | TR1 H cell development in EAE and helminth infection. (h). Each dot represents one mouse. IL-10 MFIs (average of three mice 6 a, Clinical EAE-score.Anti-CD3 was injected 35 days after MOG- s.e.m.) are reported. **P # 0.05 by Mann–Whitney U-test, two tailed. i, IFN-c/ eGFP immunization. b, Flow cytometric analysis of TH17 and exTH17 (gated on IL-10 expression of exTH17 isolated from small intestine of EAE mice. YFP1 cells) cells in dLNs. c, d, Flow cytometric analysis (c)and percentages of f–i, Representative of two independent experiments. j, Schematic of the exT 17 eGFP TR1 H cells (gated on exTH17) in dLNs (d). e, IFN-c/IL-10 expression of experiment. k, l, Flow cytometric analysis and frequencies of exTH17 and exT 17 1 exTH17 cells. a–e, Representative of three independent experiments. f, Flow TR1 H from the lung (upper panel: gated on CD4 T cells; lower panel: gated cytometric analysis of MOG-tetramer staining of intestinal CD4 cells of EAE on exTH17 cells). One experiment of three is shown. Mean 6 s.e.m., *P # 0.05, mice left untreated (control) or treated with anti-CD3. g, h, Representative flow **P # 0.005 by ANOVA (Tukey’s multiple comparison test). exT 17 6 cytometric analysis (g) and frequencies of TR1 H (gated on MOG cells)

ab c e f TGF-β + IL-6 +IL-23 IL-1β + IL-6 +IL-23 40 TH17 Control + Smad3 Inhibitor Control Smad3 inhibitor g 40 25.9% 8.9% 30 ** ** ** 21% 31% (%) s 30 17 Katushka Katushka Katushka Katushka (%) H 20 t 17 H exT 20 1 TR1exTH1 R exT IL-17A IL-17A IL-17A IL-17A T 10 10 Tr1

2% 0.6% 0 2.1% 2.3% [TGF-β] Anti TGF-β 0 [TGF-β] YFP YFP + IL-1β YFP YFP 1,000 d *** 1,500 23.6% 3.3% 800 12 23% 12% ** ** ** eGFP eGFP 439 MFI 133 MFI eGFP 1,051 MFI eGFP 486 MFI (MFI)

600 (MFI) 1,000 IL-10 IL-10 8 IL-10 IL-10 eGFP

400 eGFP 500

IL-10 200 4 IL-10 Colitis score 0 0 β β CD4 CD4 [TGF- ] Anti TGF- 0 CD4 CD4 [TGF-β] + IL-1β TH17 TR1exTH17 g h Control + FICZ + AhR Ant. i * *** * * * NS NS * NS *** Control + FICZ + AhR antagonist 50 * * ** * *** 1,400 *** *** ** * *** Click’s medium Click’s medium RPMI medium RPMI medium *** *** *** *** *** *** *** *** *** *** 23.4% 40% 9.72% +FICZ +FICZ 40 eGFP 606 MFI 1,107 MFI 364 MFI eGFP 17.9% 41% 12% 26% (%)

30 (MFI) 1,304 MFI 2,830 MFI 972 MFI 1,745 MFI 17 H IL-10 700 IL-10 exT 20 eGFP 1 R T

10 IL-10

0 0 CD4 CD4 [TGF-β] [TGF-β]

Figure 4 | TGF-b1 via Smad3, and AhR support the conversion of TH17 to presence of TGF-b1 (diluted as above), IL-6, IL-23 6 Smad3 inhibitor. exT 17 TR1. TH17 cells were differentiated in vitro in the presence of IL-6, IL-23 with f, Percentages and IL-10 MFI of TR1 H cells. Technical replicates (n 5 3) of TGF-b1 or with IL-1b and anti-TGF-b monoclonal antibody. TGF-b1 was one experiment out of five are shown. Mean and s.e.m., *P # 0.05, **P # 0.005 21 exT 17 diluted 1:2 starting from 4 ng ml . a, Flow cytometric analysis of TH17, by paired t-test. g, Flow cytometric analysis of TR1 H (gated on exTH17) exT 17 exTH17 and TR1 H (gated on exTH17). b, Percentages and IL-10 MFI of cultured in the presence of TGF-b1 (diluted as above), IL-6, IL-23 6 AhR exT 17 exT 17 TR1 H cells. Technical replicates (n 5 2) of one experiment out of seven. ligand (FICZ) or AhR antagonist. h, Percentages and IL-10 MFI of TR1 H exT 17 c, d, Endoscopic pictures and score of mice injected with TH17 or TR1 H (gated on exTH17). Technical replicates (n 5 3) of one experiment out of five cells polarized with TGF-b1. Stool inconsistency (s), increased mucosal are shown. Mean and s.e.m.; *P # 0.005, **P # 0.005, ***P # 0.0005 by granularity (g) and a lack of translucency (t). Each dot denotes one biological ANOVA (Tukey’s multiple comparison test). NS, non-significant. i, Flow exT 17 replicate. Mean and s.e.m., ***P # 0.0005 by Mann Whitney U-test, two tailed. cytometric analysis of TR1 H cells cultured in the presence of TGF-b1, IL-6, exT 17 e, Flow cytometric analysis of TH17, exTH17 and TR1 H cells cultured in the IL-23 6 FICZ in the indicated medias. One experiment of two is shown.

d Thus, conversion of TH17 into TR1 cells is a physiological mech- generation is defective in mutant Ahr mice13.Thus,theroleofAhr anism, occurring in steady state and favoured during worm and bac- expression in potentially inflammatory TH17 cells could be to enable a terial infection. switch to a regulatory fate and terminate the immune response30. We sought candidate pathways that drive TH17 conversion into TR1 Indeed, as reported28 CD4 T cells, skewed towards TH17 in the presence cells. Expression of twelve genes from our list of TH17-relevant genes of TGF-b1, express high levels of AHR (Extended Data Fig. 9d). Finally, exT 17 exT 17 appeared to be higher in both TR1 and TR1 H cells, than in TH17 to test whether AhR activation influences TH17 to TR1 H conversion cells with 9 of them being associated with TGF-b signalling (Extended we added an AhR ligand (FICZ) or AhR antagonist to the culture. FICZ exT 17 Data Fig. 8a, b). TGF-b1, in combination with IL-6 and IL-23, pro- significantly enhanced the development of TR1 H cells while the motes in vitro development of potentially pathogenic TH17 cells16,27. AhR antagonist reduced the conversion (Fig. 4g, h). Moreover, when Some of these TH17 cells expressed IL-10 (ref. 16). We found that replacing Click’s medium, which is rich in AhR ligands, with RPMI, exT 17 TGF-b1, in contrast to IL-1b, which is known to induce IL-10Negative poor in AhR ligands, the development of TR1 H was reduced. TH17 cells18,28, promoted TH17 cell plasticity and conversion to TR1 Adding FICZ to RPMI medium rescued the conversion (Fig. 4i). exT 17 cells in a dose-dependent manner (Fig. 4a, b and Extended Data These in vitro generated TR1 H cells also exhibited regulatory func- Fig. 9a). Since in vivo T cells are likely exposed to both TGF-b1 and tion (Extended Data Fig. 9e). Finally we purified intestinal TH17 cells exT 17 IL-1b simultaneously we tested TR1 H cell development in the and showed that some of these cells when re-stimulated in vitro with exT 17 presence of both cytokines and found that TR1 H cells mature TGF-b1 and FICZ convert into TR1 cells (Extended Data Fig. 9f). normally. However neutralizing TGF-b monoclonal antibody Overall our study shows TH17 cells can transdifferentiate to TR1 exT 17 impaired the development of TR1 H cells (Extended Data Fig. 9b). cells during an immune response and in the presence of TGF-b1, AhR Importantly, while TH17 cells generated with TGF-b1/IL-6/IL-23 are activation promotes this conversion. We believe that TH17 cell plas- exT 17 able to promote colitis, TR1 H cells generated under the same con- ticity might be exploited to develop new and more effective therapies ditions failed to induce disease. Thus, TGF-b1 is important for that restore immune tolerance in chronic inflammatory/autoimmune exT 17 TR1 H development and despite TGF-b1, TH17 cells remain colito- diseases without incurring the deleterious side-effects associated with genic as long as they do not convert into TR1 cells (Fig. 4c, d). current systemic immunosuppressive therapies. b Among the TGF- signalling pathway molecules, Smad3 decreases Online Content Methods, along with any additional Extended Data display items 29 RORct activity and therefore reduces TH17 cell development .We and Source Data, are available in the online version of the paper; references unique asked if TGF-b1 promotes TH17 to TR1 conversion by modulating to these sections appear only in the online paper. Smad3. Thus, when we blocked Smad3 during in vitro TH17 differ- exT 17 Received 24 September 2014; accepted 10 April 2015. entiation, induction of TR1 H cells was reduced (Fig. 4e, f and Extended Data Fig. 9c). Thus, TGF-b1, likely through Smad3, pro- Published online 29 April; corrected online 8 July 2015 (see full-text HTML version for details). motes TH17 to TR1 conversion. TH17 and TR1 cells differentiated in the presence of TGF-b1express 1. Hirota, K. et al. Fate mapping of IL-17-producing T cells in inflammatory responses. a high level of AhR13,28. AhR promotes Il10 transactivation and TR1 Nature Immunol. 12, 255–263 (2011).

2. Annunziato, F. et al. Phenotypic and functional features of human Th17 cells. J. Exp. 24. Chen, F. et al. An essential role for TH2-type responses in limiting acute tissue Med. 204, 1849–1861 (2007). damage during experimental helminth infection. Nature Med. 18, 260–266 3. Graf, T. & Enver, T. Forcing cells to change lineages. Nature 462, 587–594 (2009). (2012). 4. Komatsu, N. et al. Heterogeneity of natural Foxp31 T cells: a committed regulatory 25. Panzer, M. et al. Rapid in vivo conversion of effector T cells into Th2 cells during T-cell lineage and an uncommitted minor population retaining plasticity. Proc. Natl helminth infection. J. Immunol. 188, 615–623 (2011). Acad. Sci. USA 106, 1903–1908 (2009). 26. Milner, J. D. et al. Impaired TH17 cell differentiation in subjects with autosomal 5. Nylander, A. & Hafler, D. A. Multiple sclerosis. J. Clin. Invest. 122, 1180–1188 dominant hyper-IgE syndrome. Nature 452, 773–776 (2008). (2012). 27. Lee, Y. et al. Induction and molecular signature of pathogenic TH17 cells. Nature 6. Huber, S., Gagliani, N. & Flavell, R. A. Life, death, and miracles: Th17 cells in the Immunol. 13, 991–999 (2012). intestine. Eur. J. Immunol. 42, 2238–2245 (2012). 28. Ghoreschi, K. et al. Generation of pathogenic TH17 cells in the absence of TGF-beta 7. Roncarolo, M. G. & Battaglia, M. Regulatory T-cell immunotherapy for tolerance to signalling. Nature 467, 967–971 (2010). self antigens and alloantigens in humans. Nature Rev. Immunol. 7, 585–598 29. Martinez, G. J. et al. Smad3 differentially regulates the induction of regulatory and (2007). inflammatory T cell differentiation. J. Biol. Chem. 284, 35283–35286 (2009). 8. Huber, S. et al. Th17 cells express interleukin-10 receptor and are controlled by 30. Stockinger, B., Di Meglio, P., Gialitakis, M. & Duarte, J. H. The aryl hydrocarbon Foxp32 and Foxp31 regulatory CD41 T cells in an interleukin-10-dependent receptor: multitasking in the immune system. Annu. Rev. Immunol. 32, 403–432 manner. Immunity 34, 554–565 (2011). (2014). 9. Gagliani, N. et al. Coexpression of CD49b and LAG-3 identifies human and mouse Supplementary Information is available in the online version of the paper. T regulatory type 1 cells. Nature Med. 19, 739–746 (2013). 10. Heinemann, C. et al. IL-27 and IL-12 oppose pro-inflammatory IL-23 in CD41 T Acknowledgements The authors would like to thank C. Lieber, P. Musco, cells by inducing Blimp1. Nature Commun. 5, 3770 (2014). E. Hughes-Picard and J. Alderman for expert administrative assistance. J. Stein, 11. Okamura, T. et al. TGF-b3-expressing CD41CD252LAG31 regulatory T cells L. Evangelisti and C. Hughes for generating the IL-17A IRES-eGFP-CRE-ERT2 control humoral immune responses. Nature Commun. 6, 6329 (2015). constructs, embryonic stem cells and chimaeric mice, respectively. We thank 12. Littman, D. R. & Rudensky, A. Y. Th17 and regulatory T cells in mediating and E. Baiocchi for remote key support. LXG is supported by grants K01ES025434 from restraining inflammation. Cell 140, 845–858 (2010). NIH/BD2K and P20 COBRE GM103457 from NIH/NIGMS. E.E. was supported by the 13. Apetoh, L. et al. The aryl hydrocarbon receptor interacts with c-Maf to promote the DFG (EXC 257 NeuroCure and SFB633) and by the Crohn’s & Colitis Foundation of differentiation of type 1 regulatory T cells induced by IL-27. Nature Immunol. 11, America (#311143). N.G. is supported by the Dr. Keith Landesman Memorial 854–861 (2010). Fellowship of the Cancer Research Institute. S.H. is supported by the DFG (HU1714/3) 14. Zhou, L. et al. TGF-b-induced Foxp3 inhibits TH17 cell differentiation by and by Ernst Jung-Stiftung Hamburg and has an endowed Hofschneider-Professorship antagonizing RORctfunction.Nature 453, 236–240 (2008). from the Stiftung Experimentelle Biomedizin. This work was supported, by the Howard 15. Beriou, G. et al. IL-17-producing human peripheral regulatory T cells retain Hughes Medical Institute, by Cariplo foundation (2013-0937 to J.G. and R.A.F.), by the suppressive function. Blood 113, 4240–4249 (2009). AbbVie-Yale Collaboration (R.A.F.) and by the Francis Crick Institute (B.S.). 16. McGeachy, M. J. et al. TGF-b and IL-6 drive the production of IL-17 and IL-10 by T Author Contributions N.G. designed and performed the experiments, analysed the cells and restrain TH-17 cell-mediated pathology. Nature Immunol. 8, 1390–1397 (2007). data and wrote the manuscript. M.C.A.V., A.I., H.X. and L.B. performed the experiments and analysed the data. N.W.P. and M.R.d.Z. optimized the isolation of intestinal cells. 17. Esplugues, E. et al. Control of TH17 cells occurs in the small intestine. Nature 475, P.L-L. provided the expertise and supervised the experiments with N. brasiliensis. T.C. 514–518 (2011). performed bioinformatics analysis and L.X.G. supervised the bioinformatics analysis. 18. Zielinski, C. E. et al. Pathogen-induced human TH17 cells produce IFN-c or IL-10 C.W. provided the IL-10Thy1.1 mice. X.Z., X.P., R.F. performed and supervised the and are regulated by IL-1b. Nature 484, 514–518 (2012). extraction, amplification and library preparation for RNA sequencing. M.J.C., Y.D. and 19. Kamanaka, M. et al. Expression of interleukin-10 in intestinal lymphocytes B.B. assisted with expression profiling and data analysis. M.R.d.Z., P.L.-L., J.G., R.S.P. and detected by an interleukin-10 reporter knockin tiger mouse. Immunity 25, E.E. discussed and interpreted the results. N.W.P. discussed and interpreted the results 941–952 (2006). and helped in writing the paper. B.S. provided the IL-17A Cre mice and the protocol for 20. Ciofani, M. et al. A validated regulatory network for Th17 cell specification. Cell 151, the intracellular staining. R.A.F. and S.H. wrote the manuscript and supervised the 289–303 (2014). project. 21. Yosef, N. et al. Dynamic regulatory network controlling TH17 cell differentiation. Nature 496, 461–468 (2014). Author Information The RNA sequencing data have been deposited in the Gene 22. Cua, D. J. et al. Interleukin-23 rather than interleukin-12 is the critical cytokine for Expression Omnibus (GEO) database under the accession number GSE68242. autoimmune inflammation of the brain. Nature 421, 744–748 (2003). Reprints and permissions information is available at www.nature.com/reprints. The 23. Gause, W. C., Wynn, T. A. & Allen, J. E. Type 2 immunity and wound healing: authors declare no competing financial interests. Readers are welcome to comment on evolutionary refinement of adaptive immunity by helminths. Nature Rev. Immunol. the online version of the paper. Correspondence and requests for materials should be 13, 607–614 (2013). addressed to R.F. ([email protected]) or S.H. ([email protected]).

METHODS further digestion with collagenase from Clostridium Histolyticum (#2139 Mice. C57BL/6 Rag12/2 and Rosa26floxSTOPflox eYFP mice were purchased from SIGMA) and DNase at 37 uC for 1 h (for LPL). We then further separated cells The Jackson Laboratories. IL-17ACRE and IL-10Th1.1 (CD90.1) mice were kindly with a Percoll gradient. Unless otherwise indicated, we isolated cells from the small provided from B. Stockinger and C. Weaver respectively1,31. IL-17-IGCE intestine (duodenum, ileum and jejunum) of mice treated with antibodies to CD3. iFate mice, Foxp3RFP, IL-10eGFP, IL-17AKatushka, IFN-cKatushka reporter mice and Flow cytometry antibodies and intracellular cytokine staining. We stained CD4dnIL-10Ra (DNIL-10R) were generated and breed in our laboratory. reporter mouse T cells with monoclonal antibodies to CD4 (GK1.5, Cat # 100428 or mice were purchased from Jackson Laboratory. The Fate1 mice results from the RM4-5 Cat # 100536), CD8 (53-6.7 Cat # 100722), NK1.1 (RM4-5 Cat # breeding of the original Foxp3RFP IL-10eGFP IL-17AKata mice9,17,19 with IL-17ACRE 100536), CD19 (6D5 Cat # 115508), CD11b (M1/70 Cat # 101216), CD11c R26YFP mice1. In this model, only high level of Il17a transcription induces the (N418 Cat # 117318), cdTCR (GL3 Cat # 118123), CD210 (BD Bioscience, Cat expression of Cre recombinase, which deletes the stop sequence 59 to YFP. In this # 559914), LAG-3 (C9B7W Cat # 125209), CD49b (HMa2 Cat # 103506) and mouse, cells that have previously expressed high level of Il17a, delete the stop CCR6 (29-2L17 Cat # 129817), all antibodies expected where indicated are pur- cassette and are thus permanently marked by the expression of YFP. Importantly it chased from eBiolegend. Importantly, CD49b and LAG-3 staining were per- has been previously described that this IL-17A fate reporter allele faithfully marks formed at 37 uC for 45 min. Although in the figure legends we referred only to 1 TH17 cells that have acquired full effector function1. Of note, we observed that all CD4 T cells, in each FACS related experiment and FACS-sorting experiment we IL-17A bright cells are YFP1, whereas IL-17A dim cells remain YFP negative, have specifically analysed CD41 T cells CD82, NK1.12, CD192, CD11b2, confirming the data already published whereby only IL-17A high expressing cells, CD11c2, cdTCR2. For intracellular cytokine staining the cells were re-stimulated 21 fully differentiated TH17 cells, are permanently marked with YFP1. for 3 h at 37 uC with phorbol 12-myristate 13-acetate (PMA) (Sigma, 50 ng ml ) To generate the DNIL-10R FATE, the CD4-DNIL-10R were crossed with and ionomycin (Sigma, 1 mgml21) in the presence of Golgistop (BD Bioscience). IL-17ACRE R26YFP IL-10eGFP Foxp3RFP. Cells were then fixed in paraformaldehyde for 20 min at room temperature. After All mice were kept under specific pathogen-free (SPF) conditions in the animal washing, the cells have been permeabilized (NP40) and stained at 4 uC with anti- facility at Yale University. We used age- and sex-matched littermates between 12 IL-17A (TC11-18H10.1 Cat # 506925), anti-IFNc (BD Bioscience, Cat # 554412), and 20 weeks of age. Animal procedures were approved by the Institutional anti-IL-4 (BD Bioscience, Cat # 554435) and anti-Rorct (BD Bioscience Cat # Animal Care and Use Committee of Yale University. Both female and male mice 553178) antibodies for 30 min. Lymphocytes were re-suspended in PBS, 0.5% FBS, were used in experiments. Wherever possible, preliminary experiments were per- 5 mM EDTA and acquired with an LSRII cytometer (BD Bioscience). exT 17 1 RFP- formed to determine requirements for sample size, taking into account resources In vitro TR1 H cell differentiation. We FACS sorted CD4 Foxp3 available and ethical use. Exclusion criteria such as inadequate staining or low cell IL-17AKatushka- IL-10eGFP- R26YFP- cells with FACSAria II Cell Sorter (BD yield due to technical problems were pre-determined. Animals were assigned Biosciences) and activated them with plate-bound monoclonal antibodies randomly to experimental groups. Each cage contained animals of all the different to CD3 (10 mgml21, 145-2C11) and CD28 (1–2 mgml21, PV-1) in the presence experimental groups expected for the mice treated with antibiotic. of mouse recombinant TGF-b (0.25–4 ng ml21), IL-6 (20 ng ml21), IL-23 Generation of inducible Fate mice. The IL-17A IRES-eGFP-CRE-ERT2 mice (20 ng ml21), and antibodies to IFN-c (XMG1.2, 10 mgml21) and IL-4 (11B11, (iFate) mice were generated following the same targeting strategy used previously 10 mgml21). When specified, IL-1b (50 ng ml21), antibodies to TGF-b (5 mgml21, eGFP 17 to generate the IL-17A mice . Briefly, a cassette encoding for a fusion protein 1D11,) and FICZ (100 nM; Enzo Life Sciences), Smad3 inhibitor (SIS3, 3 mM, consisting of the Internal Ribosome Entry Site (IRES), eGFP, Cre and human EDM Millipore)32 or AhR antagonist (10 mM, EDM Millipore), were added to the modified Oestrogen Receptor (ERT2) (IGCE) was linked to a Frt-flanked neo- culturing media. All cytokines were purchased from R&D. Click’s (Irvine mycin (NEO) encoding cassette. The IGCE-NEO construct was cloned into a Scientific) or RPMI (SIGMA-ALDRICH) (when indicated) media were supple- plasmid containing two homology arms on the Il17a gene: the 59 homology mented with 10% FBS, L-glutammine (2 mM), penicillin (100 U ml21) and arm corresponds to a 4,445-bp fragment of the Il17a gene to the fourteenth base b-mercaptoethanol (40 nM). After 4–5 days of culture, the cells were acquired pairs after the stop codon of the gene using an Asc cloning site while the 39 at the FACS. homology arm of the targeting construct consists of the genomic sequence of Foxp3RFP IL-17AKatushka IL-10eGFP triple reporter were injected with anti-CD3 3,258 bp spacing from fifteenth base pairs after the stop codon of the Il17a gene mAb and 12 h after the first injection or 4 h after the third injection a pure using a Not cloning site. 1 RFP2 Katushka+ eGFP2 population of intestinal CD4 Foxp3 IL-17A IL-10 cells were Drug-resistant ES cell clones were screened for homologous recombination by FACS sorted and restimulated in vitro in the presence of irradiated splenocytes PCR. To obtain chimeric mice, correctly targeted ES clones were injected into (1:4 ratio). The cells were stimulated for 5 days in the presence of soluble anti-CD3 C57BL/6 blastocysts, which were then implanted into CD1 pseudopregnant foster monoclonal antibody (2 mgml21), IL-6 (20 ng ml21) and where indicated anti- mothers. Male chimaeras were bred with C57BL/6 to screen for germline trans- TGF-b (5 mgml21), TGF-b (0.25 ng) and FICZ (100 nM). mitted offspring. Germline transmitted mice were bred with germline Flippase RNA amplification, extraction and sequencing. We isolated intestinal lympho- expressing transgenic mice to remove the neomycin gene. cytes from two independent experiments, each using 5 mice injected with anti- Mice bearing the construct were screened by PCR and bred with germline Flp CD3 monoclonal antibody. The cell populations indicated in Fig. 2 were FACS- expressing transgenic mice to remove the neomycin gene. After removal of the NEO cassette, IL-17A iFate mice were crossed with R26eYFP and all of the cells that sorted from these two independent experiments and the cells of each population actively express IL-17A were eGFP1 but still YFP1 negative. However, unlike in were pooled before the RNA extraction, amplification and sequencing. Around Fate1 mice, IL-17A expressing cells become permanently marked as YFP1 after 5,000 cells for each population were processed. After sorting, the cells were washed treatment with tamoxifen as ERT2 sequesters the Cre in the cytoplasm until 2 times (1,500 r.p.m., 2 min, 4 uC) with 1 ml phosphate-buffered saline (PBS) and tamoxifen binds to ERT2. finally suspended in 2.5 ml PBS (containing 0.5 ml RNaseOut (Invitrogen) and 0.5ml The efficiency of CRE-mediated recombination after tamoxifien is reported as dithiothreitol (DDT) (Invitrogen)). After that, the cytoplasm RNA was isolated as 33,34 frequencies of YFP1 cells (YFP) in Fig. 1d and Extended Data Figs 4, 6 and 7 and is described previously . Briefly, 2.5 ml of 2x selected cytoplasm lysis buffer (SCLB) the result of the following calculation: YFP1 cells (Gate213) / (IL-17AeGFP1 YFP2 was added and the cells were lysed by pipetting up and down for 5 times. The entire cells (Gate 1) 1 YFP1 cells(Gate213). lysate solution was spun at 8,000 r.p.m. for 5 min in a chilled centrifuge (4 uC) and Anti-CD3, antibiotic and tamoxifen treatments. Mice were injected with anti- the supernatant (,5 ml), which contained the total cytoplasm RNA, was transferred CD3 mAb (2C11,15–50 mg per mouse) intra-peritoneally two times every other to a PCR tube strip with individually attached dome caps (USA Scientific). The day. Usually the mice were sacrificed 4 h after the last injection, unless differently mRNA selection, reverse transcription and cDNA amplification was performed as 33,34 indicated. Vancomycin was dissolved in water to a final concentration of 0.5 g l21 described previously with some modifications .Briefly,59-phosphorylated and administered in drinking water for 4 weeks before starting the experiment. oligo-GdT24 (pGdT24) primer was used to selectively reverse transcribe Tamoxifen (Sigma) was dissolved in corn oil (Fluka, Sigma) to a final concentra- mRNAs. The first strand cDNA was synthesized with Superscript Reverse tion of 20 mg ml21. Mice were injected with tamoxifen (4 mg each) one day before Transcriptase III (Invitrogen). Then, the double-stranded cDNA (dscDNA) was each anti-CD3 monoclonal antibody injection or as depicted in the scheme of the generated. The cDNA was purified with the Genomic DNA Clean & Concentrator other experiments. To avoid to interfere with the effect of anti-CD3 mAb and/or kit (Zymo). Afterwards, several steps including DNA end-blunting, 59-end phos- the migration of cells to the intestine, the oil + tamoxifen was injected subcuta- phorylation and ligation were performed with The End-It DNA End-Repair Kit neously. Only for the experiments involving N. brasiliensis infection we injected (Epicentre) with T4 DNA ligase (Epicentre). The product was directly amplified tamoxifen intraperitoneally (i.p.). using REPLI-g UltraFast Mini Kit (QIAGEN) and purified using the same Lymphocyte isolation from small intestine. After removal of the Peyer’s patches, Genomic DNA Clean & Concentrator kit (Zymo) column. Finally, 3 to 5 mg(up we isolated intraepithelial lymphocytes (IELs) and lamina propria lymphocytes to 8 mg) of amplified cDNA derived from mature mRNA was obtained, which was (LPLs) by incubation with 1 mM DTE at 37 uC for 30 min (for IEL), followed by then evaluated by PCR and fragmented to construct sequencing library.

We follow the standard Illumina HiSeq2000 protocol to make the library. older than 7 weeks. Mice were killed 3–4 days after the injection, at a time when Briefly, the amplicon was fragmented to an approximately 200–500 bp size range they displayed severe clinical symptoms of sepsis (weight loss, dehydration, leth- exT 17 by a Bioruptor Sonicator (Diagenode). After purification with DNA Clean & argy) and the presence of TH17, exTH17 and TR1 H cells was tested in the small Concentrator kit (Zymo), end-repairing, 39-A tailing and ligation with adaptor intestine by FACS. 1 was performed. Then, a 50 bp range DNA (250–300 bp) was selected by gel TH17 transfer colitis, endoscopic and histologic analysis. CD4 were isolated electrophoresis (E-gel EX 2%, Invitrogen) and barcode added by PCR using from the IL-17AeGFP Foxp3RFP double report mice and cultured with irradiated Phusion High-Fidelity DNA polymerase (NEB) for 8 cycles. The product was size antigen presenting cells (1:4 ratio), in the presence of soluble anti-CD3 mAb selected again and the DNA concentration was quantitated by a Bioanalyzer (2C11, 1 mgml21), IL-6 (20 ng ml21), IL-23 (50 ng ml21) and TGF-b1 (0.25 ng (Agilent). Multiple samples were mixed and loaded to the Hi-Seq2000 for sequen- ml21) along with neutralizing antibodies for IFN-c (XMG1.2 clone, 10 mgml21) cing performed with 50 bp single-end reads. and IL-4 (11B11 clone, 10 mgml21). After 5 days of in vitro culture, a pure 1 RNA-Seq data were aligned to the Mus musculus GRCm37 genome using population of 10,000 pathogenic (p) TH17 cells (FACS sorted as CD4 IL- 35 Tophat2 and default settings . Duplicate reads were removed with samtools 17eGFP1 Foxp3RFP2) were injected intra peritoneally into Rag12/2 mice at 1:1 36 rmdup command . Count data was generated using HTSeq-count and FPKM ratio with the following cell populations: TH17, exTH17, TR1 exTH17 and TR1 cells. 37,38 data was generated with Cufflinks . To determine genes important for the TR1 These later populations were FACS sorted from the intestine of Fate1 4 h after the exTH17 conversion we performed a differential expression test using DESeq2 second injections of anti-CD3 monoclonal antibody. exT 17 comparing TH17 cells with TR1 exTH17 and TR1 cells. To determine which cell When indicated (Fig. 4c, d) both TH17 and TR1 H cells were generated in populations were more closely related, pearson correlation values between samples vitro under the same condition (CD3 (10 mgml21) and CD28 (1–2 mgml21); TGF- were calculated on log2 FPKM data after a maximum filter of FPKM .1, based on b1 (0.25–0.5 ng ml21), IL-6 (20 ng ml21), IL-23 (20 ng ml21), and antibodies the subset of genes listed in Supplementary Table 1 and 2. In order to determine to IFN-c (10 mgml21) and IL-4 (10 mgml21)) and then sorted and transferred the grouping of cell populations, hierarchical cluster analysis was performed on (n 5 10,000 to 20,000 cells) into Rag-12/2 mice. the distances based on correlation or gene expression. The linkage criteria com- exTH17 When indicated (Extended Data Fig. 9), the TR1 cells were generated in plete was used for clustering analysis. vitro in the presence of TGF-b1(1ngml21), IL-6 (20 ng ml21), IL-23 (20 ng ml21), Real time PCR. 21 21 Total RNA was extracted from cells using TRIzol reagent. To and antibodies to IFN-c (XMG1.2, 10 mgml ) and IL-4 (11B11, 10 mgml )and synthetize the cDNA we then used the High Capacity cDNA Sythesis Kit (Applied FICZ (100 nM; Enzo Life Sciences), FACS sorted and transferred (n 5 10,000 cells) Biosystem) and RT–PCR was performed using the TaqMan Fast Universal PCR 2/2 into Rag1 mice at 1:1 ratio with pTH17 cells. Master Mix and TaqMan Gene Probes (Applied Biosystems) on a 7500 Fast Real- Colonoscopy was performed in a blinded fashion using the Coloview system time PCR system machine (Applied Biosystem). Samples were run in duplicate or (Karl Storz, Germany). Briefly, colitis score was addressed considering the con- triplicate and expression levels were calculated as relative to the expression of sistence of stools, granularity of the mucosal surface, translucency of the colon, endogenous HPRT or Polr2a. fibrin deposit and vascularization of the mucosa (0–3 points for each parameter). Experimental autoimmune encephalomyelitis and tetramers staining. Mice Haematoxilin and eosin staining were performed on paraffin sections of colon were immunized sub-cutaneously with an emulsion of 250 mg of MOG pep- 35-55 previously fixed in Bouin’s fixative solutions. tide (Yale Keck facility) and CFA (BD Difco). At the time of immunization and 48 h after, mice received 200 ng pertussis toxin (PTx, List Biological Laboratories) Statistical analysis and FACS analysis. Statistical analysis were performed using per each injection. The clinical score of EAE development was addressed daily Prism 5.0 (Graphpad Software) Paired t test, Non parametric Mann–Whitney according to guidelines: 0, no signs of disease; 0,5, tail weakness; 1, complete tail U-test, ANOVA (post test Tukey, Bonferroni or Dunnet) were used according paralysis; 2, partial hind limb paralysis; 2,5, unilateral complete hind limb para- to the type of experiments. Log10-transformed values for cell counts were used in lysis; 3, complete bilateral hind limb paralysis; 3,5, complete hind limb paralysis Fig. 1. P-values # 0.05 were considered significant (*: P , 0.05; **: P , 0.005; ***: and partial forelimb paralysis; 4, total paralysis of forelimbs and hind limbs, P , 0.0005); P values .0.05; non-significant (NS). All flow cytometry data have moribund. All mice experiments were conduced according to IACUC policies. been analysed with FlowJo (Treestar). No statistical methods were used to pre- determine sample size. To identify MOG38-49 (mouse myelin oligodendrocyte glycoprotein 38-49, ‘‘GWYRSPFSRWH’’) specific T-cells, 107 cells per ml were incubated with neaur- 21 31. Maynard, C. L. et al. Regulatory T cells expressing interleukin 10 develop from aminidase (0.5 U ml , neuraminidase type X from Clostridium perfringens, Foxp31 and Foxp32 precursor cells in the absence of interleukin 10. Nature Sigma) in serum-free DMEM at 37 uC/5% CO2 for 25min. After this, the cells Immunol. 8, 931–941 (2007). were stained with the MOG38-49/I-A(b)-tetramer allophycocyanin (APC)-labelled 32. Jinnin, M., Ihn, H. & Tamaki, K. Characterization of SIS3, a novel specific inhibitor (NIH Tetramer Facility) for 4 h, at room temperature in DMEM, 2%FBS. Cells of Smad3, and its effect on transforming growth factor-beta1-induced were then stained for surface antigens and acquired at the FACS. extracellular matrix expression. Mol. Pharmacol. 69, 597–607 (2006). Nippostrongylus brasiliensis infection and isolation of lymphocytes from the 33. Guo, S. et al. Nonstochastic reprogramming from a privileged somatic cell state. Cell 156, 649–662 (2014). lung. Third-stage larvae (L3) of N. brasiliensis were recovered from coprocultures of 34. Pan, X. et al. Two methods for full-length RNA sequencing for low quantities of infected mice. We infected mice by injecting subcutaneously 625 parasites in 0.2 ml cells and single cells. Proc. Natl Acad. Sci. USA 110, 594–599 (2013). PBS at the base of the tail, as previously described9.Micewereeuthanizedatdifferent 35. Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence of time points, as described, and lymphocytes were isolated from lungs. Cell suspensions insertions, deletions and gene fusions. Genome Biol. 14, R36 (2013). from lungs were obtained digesting the organs, previously cut in small pieces, in 10% 36. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics FBS RPMI media in the presence of DNase and collagenase D, as previously 25, 2078–2079 (2009). 9 37. Trapnell, C. et al. Transcript assembly and quantification by RNA-Seq reveals described . After digestion, cell suspensions were processed onto a Percoll gradient unannotated transcripts and isoform switching during cell differentiation. Nature (40% on 100%) and lymphocytes were then processed for FACS analysis. Biotechnol. 28, 511–515 (2010). 1 2 Staphylococcus aureus infections. S. aureus (ATCC 14458, SEB TSST-1 ) was 38. Anders, S., Pyl, P. T. & Huber, W. HTSeq-a Python framework to work with high- injected intravenously (1010 colony-forming units per mouse) into Fate1 mice not throughput sequencing data. Bioinformatics 31, 166–169 (2014).

Extended Data Figure 1 | Description of Fate1 mice and characterization of pooled cells from at least 7 mice. Mean and s.e.m., ***P # 0.0005 by ANOVA eGFP1 exTH17 IL-10 cells under steady state condition. a, Constructs (Tukey’s multiple comparison test). c–e, Under steady state conditions, contained in the Fate1 mice. b, During anti-CD3 mAb induced transient intestinal lymphocytes were isolated and re-stimulated in vitro for 3 h with eGFP1 inflammation in the S.I., a sufficient number of exTH17 IL-10 was PMA/ionomycin for the intracellular staining of IFN-c and IL-4, while they were generated to test whether Fate1 mice faithfully report IL17A and IL-10 freshly analysed for the expression of CCR6 and RORct. c, Frequencies of IFN-c 1 Katushka– – eGFP1 expression. In particular, TR1 (CD4 IL-17A YFP IL-10 and IL-4 among the exTH17 cells is shown. Pie chart reports the frequencies of RFP– 1 Katushka1 1 eGFP– RFP– eGFP1 Foxp3 ), TH17 (CD4 IL-17A YFP IL-10 Foxp3 ), the indicated cytokine among the exTH17 IL-10 . One biological replicates eGFP1 1 Katushka– 1 eGFP1 RFP– exTH17 IL-10 (CD4 IL-17A YFP IL-10 Foxp3 ) and out of five is shown. d, e, Frequencies and MFI of CCR6 (d)andMFIofRORct 1 Katushka– 1 eGFP– RFP– 1 Katushka– – eGFP1 1 exTH17 (CD4 IL-17A YFP IL-10 Foxp3 ) were FACS are reported for TR1(CD4 IL-17A YFP IL-10 ), TH17 (CD4 IL- 1 Katushka1 1/– eGFP– 1 1 Katushka– 1 sorted from the small intestine of anti-CD3 monoclonal antibody treated Fate 17A YFP IL-10 ), exTH17 IL-10 (CD4 IL-17A YFP eGFP1 mice and mRNA expression relative to TH17 cells for Il17a and relative to TR1 IL-10 )(e). Each dot represents one biological replicate. Mean and s.e.m., eGFP1 dim/high low for Il10 is reported. ExTH17 IL-10 express Il10 and Il17a .Data **P # 0.005, ***P # 0.0005 by ANOVA (Dunnett’s multiple comparison test, are cumulative of three independent experiments. In each experiment we comparison all columns vs control (TH17 cells)).

exT 17 1 1 Katushka1 RFP– 1 GFP1 RFP– Extended Data Figure 2 | Characterization of TR1 H cells. TR1 (CD4 e,TH1 (CD4 IFN-c Foxp3 ), TH2 (CD4 IL-4 Foxp3 ), Katushka– – eGFP1 1 Katushka1 1/– 1 Katushka– eGFP1 RFP– 1 IL-17A YFP IL-10 ), TH17 (CD4 IL-17A YFP IL- TR1 (CD4 IL-17A IL-10 Foxp3 ), TH17 (CD4 IL- eGFP– exT 17 1 Katushka– 1 eGFP1 Katushka1 RFP– exT 17 1 Katushka– 1 eGFP1 10 ), TR1 H (CD4 IL-17A YFP IL-10 ) were isolated 17A Foxp3 ), TR1 H (CD4 IL-17A YFP IL-10 1 RFP– 1 Katushka– 1 eGFP– RFP– from the small intestine of anti-CD3 monoclonal antibody treated-Fate mice Foxp3 ) and exTH17 (CD4 IL-17A YFP IL-10 Foxp3 ) and analysed by FACS. a–d, Frequencies of LAG-3 (a), CCR6 (b), CD49b were FACS sorted from the small intestine of anti-CD3 monoclonal antibody (c) and MFI of RORct(d) are reported. Each dot represents one biological treated mice. mRNA expression relative to HPRT of Ifng, Il2, Tbx21, Il4 and replicate. Mean and s.e.m., **P # 0.005, ***P # 0.0005 by ANOVA (Dunnett’s Gata3 of the indicated populations is reported. Mean and s.e.m., *P # 0.05, by multiple comparison test, Comparison all columns vs control (TH17 cells)). Mann–Whitney U-test, two tailed.

exT 17 1 – 1 eGFP1 Extended Data Figure 3 | TH17 fate in DNIL-10R Fate mice. Fate (WT) and exTH17 cells. c, Ratio between TR1 H (CD4 IL-17A YFP IL-10 – exT 17 1 – 1 eGFP– 1 Fate dominant negative IL-10R (DNIL-10R) were injected with anti-CD3 IFN-c ) and TH1 H (CD4 IL-17A YFP IL-10 IFN-c ) in WT and monoclonal antibody and the fate of small intestinal TH17 cells analysed. DNIL-10R mice. Mean 6 s.e.m. Each dot represents one biological replicate. *P 1 a, Frequencies of TH17 and exTH17 gated on CD4 T cells are reported. # 0.05; **P # 0.005 by Mann–Whitney U-test, two tailed. b, Representative flow cytometric analysis of the IL-10 and IFN-c expression in

Extended Data Figure 4 | Constructs and validation of iFate mice. indicated populations is reported. The mRNA expression is normalized to a, Targeting strategy and constructs of IL-17AeGFP-CRE-ERT2 mice (iFate). Black HPRT. One representative experiment out of two is shown. d, Representative circles represent Flp recombinase target (FRT) sites. b, Schematic of TH17 cell flow cytometric analysis of intestinal TH17 cells under steady state condition or development in iFate showing that TH17 cell plasticity can be tested only after after anti-CD3 monoclonal antibody (aCD3) in the absence of tamoxifen tamoxifen treatment. c, iFate mice were injected with tamoxifen and anti-CD3 treatment. The efficiency of CRE-mediated recombination after tamoxifen is GFP– – 1 1 monoclonal antibody and non TH17 cells (IL-17A YFP ), TH17 YFP cells reported as frequencies of YFP cells (YFP) and is the result of the following eGFP1 – eGFP1 1 1 eGFP1 2 1 (IL-17A YFP ), TH17 cells (IL-17A YFP ) and exTH17 cells calculation: YFP cells (gate 2 1 3) / (IL-17A YFP cells (gate 1) 1 YFP (IL-17AeGFP– YFP1) were FACS sorted. Il17a mRNA expression in the cells (gate 2 1 3).

eGFP– RFP1 1 1 1 Katushka– – Extended Data Figure 5 | Relative expression of Il17a, Il10 and Foxp3 and 10 Foxp3 ) and Foxp3 TReg IL-10 (CD4 IL-17A YFP IL- eGFP1 RFP1 1 FPKM values of signature genes of bona fide TH17 and TR1 cells. a,TR1 10 Foxp3 ) cells were isolated from the small intestine of Fate mice 1 Katushka– - eGFP1 RFP– 1 (CD4 IL-17A YFP IL-10 Foxp3 )TH17 (CD4 IL- after anti-CD3 monoclonal anttibody injections. The transcriptome of these Katushka1 1/2 eGFP1/– RFP– exTh17 1 17A YFP IL-10 Foxp3 ), TR1 (CD4 IL- populations was sequenced and the relative FPKM expressions of Il10, Il17a Katushka– 1 eGFP1 RFP– 1 Katushka– 1 17A YFP IL-10 Foxp3 ), exTh17 (CD4 IL-17A and Foxp3 compared to TR1, TH17 and Foxp3 TReg cells are reported. 1 eGFP– RFP– 1 1 Katushka– – YFP IL-10 Foxp3 ) Foxp3 TReg (CD4 IL-17A YFP IL- b, FPKM values of the indicated populations of the reported genes are shown.

exT 17 Extended Data Figure 6 | TR1 H cell development in EAE and during T cells isolated from the lung of iFate mice before (steady state) and after the helminth infection using iFate mice. a, Schematic of the experiment, showing second infection 6 tamoxifen (control N. brasiliensis (no tamoxifen) and N. iFate1 mice immunized with MOG, treated for 3 times with tamoxifen and brasiliensis (1 tamoxifen) respectively)). Cumulative dot plots of 3 biological then injected with anti-CD3 monoclonal antibody 70 and 72 days after MOG replicates are shown. One representative experiments out of 3 is shown. The immunization. The intestinal lymphocytes were analysed 4 h after the second YFP1 percentages (YFP) shown on the dot plots report the efficiency of injection of anti-CD3 monclonal antibody. b, Representative flow cytometric tamoxifen-induced CRE-recombination. The frequencies within the 1 exT 17 analysis of TH17 and exTH17 (gated on YFP cells) and TR1 H cells (gated cumulative dot/density plot report the percentage of TH17 and exTH17 among 1 1 1 on exTH17). The YFP percentages (YFP) shown on the dot plots report the the YFP cells, and the frequency of IL-10 cells among the exTH17 cells. exT 17 efficiency of tamoxifen-induced CRE-recombination. Three representative e, Frequencies of TR1 H cells (gated on exTH17). Each dot represents a biological replicates out of six are shown. c, Schematic of the experiment, biological replicates. Results are cumulative from three independent showing iFate mice infected with N. brasiliensis and injected i.p. with tamoxifen experiments. at the indicated time points. d, Representative flow cytometric analysis of CD41

eGFP Extended Data Figure 7 | Conversion of TH17 cells into TR1 over the course IL-10 expression of exTH17 cells. Pie chart reports the frequencies of the 1 exT 17 of S. aureus infection using Fate and iFate mice. a, Fate mice were left indicated cytokine among the TR1 H cells. a–c, One representative untreated (control) or injected i.v. with S. aureus (S. aureus). Representative biological replicate out of three is shown. One representative experiment out of 1 flow cytometric analysis of intestinal TH17 and exTH17 (gated on CD4 two is shown. d, Representative flow cytometric analysis of intestinal RFP– exT 17 1 Katushka– 1 eGFP1 Foxp3 ) and TR1 H cells (CD4 IL-17A YFP IL-10 ; gated lymphocytes isolated from iFate mice 4 days after S. aureus infection. One 1 on exTH17) are shown. One representative experiment out of three is shown. representative biological replicate out of 5 is shown. The YFP percentages b, Frequencies and numbers of the indicated population in the small intestine of (YFP) shown on the dot plots report the efficiency of tamoxifen-induced exT 17 1 eGFP– untreated (control) and infected mice (S. aureus) are reported. Results are CRE-recombination. e, Frequencies of TR1 H cells (CD4 IL-17A 1 CD90.11 cumulative from three independent experiments. Mean and s.e.m., **P # YFP IL-10 ; gated on exTH17) are shown. Results are cumulative from 0.005, ***P # 0.0005 by Mann–Whitney U-test, two tailed. c, IFN-c and two independent experiments.

exT 17 Extended Data Figure 8 | Gene expression of TR1, TR1 H and TH17 cells. with anti-CD3 monoclonal antibody is shown. Values shown are relative to exT 17 a, Heat map of genes selectively expressed in both TR1 H and TR1 compared TH17 cell gene expression. Mean and s.e.m. of biological independent to TH17 cells. The bioinformatics analysis is based on the genes listed in experiments (IRF8 n 5 2; SMAD3 n 5 4; FOXO1 n 5 2; STAT5a n 5 3; Supplementary Table 1. Red squares highlight genes linked to TGF-b1 SMAD4 n 5 4) except for IL-23 and Runx1 (n 5 2 technical replicates) are signalling. b, Relative mRNA expression of the indicated genes in TR1, shown. In each experiment we pooled intestinal lymphocytes isolated from 7 exT 17 1 TR1 H and TH17 cells FACS sorted form the intestine of Fate mice treated treated mice before FACS sorting.

exT 17 Extended Data Figure 9 | Characterization of in vitro generated TR1 H TGF-b1, IL-6, IL-23 or IL-1b, IL-6, IL-23. The expression is normalized to 1 cells. a,IL-1b counteracted TH17 plasticity. IL-17A MFI in TH17 (CD4 HPRT. One experiment out of two is shown. Two technical replicates are RFP– Katushka1 1/– eGFP1/– exT 17 Foxp3 IL-17A YFP IL-10 ) differentiated in the presence reported. e,TR1 H cells were polarized in vitro in the presence of TGF- 2/2 of TGF-b1, IL-6, IL-23 or IL-1b, IL-6, IL-23. One experiment out of five is b11IL-61IL-231FICZ and transferred into Rag1 mice 6 (p)TH17 cells. shown. Two technical replicates are reported. b, Dose-response effect of TGF- Endoscopic colitis score and percentage of initial body weight in the indicated exT 17 b1 on the induction of TR1 H cells cultured in the presence of IL-6, IL-23, IL- groups are shown. Each dot represents one mouse. Results are cumulative from 1b. In the last conditions we added anti-TGF-b1 monoclonal antibody. TGF-b1 three independent experiments. Mean and s.e.m., *P # 0.05 by Mann–Whitney 21 was diluted 1:2 starting from the concentration of 4 ng ml . One experiment U-test, two-tailed. f,TH17 cells were isolated from the intestine of anti-CD3 out of five is shown. Two technical replicates are reported. c, In line with the monoclonal antibody and then restimulated in vitro in the presence of either literature29, Smad3 chemical inhibition also favours TH17 cell development. anti-TGF-b1IL-6 or IL-61TGF-b11FICZ for 5 days. Frequencies of exT 17 Frequency of TH17 cells cultured in the presence or in the absence Smad3 TR1 H cells among total cells (left) and among exTH17 cells (right) are inhibitor at the indicated different concentrations of TGF-b1 (4–0.25 ng ml21). reported. Results are cumulative from three independent experiments. Each dot One experiment out of five is shown. Three technical replicates are reported. represents a pool of TH17 cells isolated from five mice treated with anti-CD3. d, mRNA expression of Ahr in CD4 T cells cultured in the presence of either Mean and s.e.m., *P # 0.05, by Mann–Whitney U-test, two tailed.